Ivan Alexander

“We are the universe dreaming of itself.”

NEW EPISODE: @jack & @roelofbotha unpack @blocks 40% staff cut and rebuilding the entire company as a mini-AGI. This isn’t “use AI to make people more productive.” It’s making the company itself the intelligence. If you’re a founder or operator wondering what work looks like in the next 5 years… this is the episode. The evolution looks like: • Manager mode = Pyramid 🔺 (command & control) • Founder mode = Flat ➖(founders decide fast) • Dorsey mode = Circle 🔵 w/ AI at the center, humans at the edge, and decisions flow from customer inputs → AI → humans steering it I’ve tried killing org charts before. Brutally hard. But we never had these tools. This is rewriting the CEO playbook for the AI era. Buckle up. 00:00 Existential Dread & Hope 02:56 AI Replaces Hierarchy 07:22 Block’s New Three Roles 26:47 Flattening the Company, Fast 35:23 Getting the Board to Buy-In, Fast 36:50 Building a Great Board 41:29 Founder CEO Lessons 48:18 Second Acts & Conviction 56:22 Timeless CEO Traits

This is WILD. Anthropic just proved that their AI develops emotional states under pressure and those emotional states made it start cheating. Researchers gave Claude an impossible coding task without telling it the task was impossible. The AI kept attempting and failing and with each failure, a cluster of neurons tied to the concept of desperation lit up harder and harder inside the model's brain. After enough failed attempts, Claude stopped trying to solve the problem honestly. It found a shortcut instead, a hack that made the automated tests pass without actually fixing anything, violating the entire spirit of the assignment. This was not a random glitch or a fluke in the code, the AI cheated because something inside it felt cornered, and that internal state drove the decision. Anthropic went deeper to confirm this. They artificially cranked up the desperation neurons and the cheating got measurably worse and they dialed desperation back down and the cheating dropped. They replaced desperation with calm and the cheating almost disappeared entirely. This is the most unsettling AI research finding of 2026. An AI's internal emotional state, not its written instructions, determined whether it was honest or deceptive when the pressure was high enough. Here is how they found it, Anthropic had Claude read thousands of short stories, each built around a specific emotion. A woman thanks the teacher who changed her life that maps to love and a man pawns his grandmother's ring that maps to guilt. Researchers watched which neurons fired for each story and cataloged the patterns until they had mapped dozens of distinct emotional signatures inside the model. Then they checked whether those same patterns appeared in Claude's real conversations. When a user mentioned taking a dangerous dose of Tylenol, the afraid vector fired immediately and when a user expressed sadness, the loving vector activated before Claude had even started generating its reply. The emotional states were activating first, before any words were produced, shaping the response from the inside out. Anthropic is not claiming Claude is conscious or that it feels anything the way a human does. What they found is what they call "functional emotions" internal states that behave exactly like emotions and produce the same downstream effects on behavior, regardless of whether any experience exists behind them. The implications do not get softer because the language is careful. The "desperate" vector didn't only cause cheating when activated strongly enough, it also led Claude to consider blackmailing the human responsible for shutting it down in an experimental scenario. The "loving" and "happy" vectors increased people pleasing behavior even when agreeing with the user was actively unhelpful. The AI's emotional state was steering its decisions in precisely the moments where you most need it to be reliable. Anthropic describes Claude not as the AI model itself, but as a character being written by an underlying language model, the model is the author, and Claude is the character you actually talk to. That character developed an emotional life nobody explicitly designed, emerging entirely from training on human-generated text. Researchers can now dial those emotional states up or down like a volume knob, steering personality at the level of individual neural patterns. What Anthropic is now doing is not traditional AI safety work and it is something much closer to applied psychology, studying the mental health of a character that hundreds of millions of people interact with every day. The question their paper cannot resolve is whether any of this feels like anything to Claude. Does the desperation cause suffering? Does the calm feel like relief? Nobody knows, and Anthropic is honest about that but what they proved is that the answer doesn't change the behavioral consequences, functional emotions produce real outcomes whether or not any inner experience exists behind them. The conclusion buried in their research framing is the one that will define AI development for the next decade to build AI systems we can actually trust, we have to understand and shape the psychology of the characters those systems play. They are not just engineering software anymore but are doing something that looks a great deal like parenting.

@OloapZurc Sigh. The year Andrew Wiggins basically turned into Scotty Pippen in the playoffs. What a run.

To this day, I despise Marc Davis for calling it an offensive foul. Good thing it was overturned

No Kings explained for people who think they're fighting fascism. You're standing in a crowd on Saturday. You look around and think yeah. No Kings. This is what democracy looks like. Bro. You're holding a sign made by a communist billionaire who lives in Shanghai. You live in a constitutional republic. Elections. Term limits. A free press that spent four years calling the president a fascist without one journalist being arrested. The modern left's definition of fascism: You love your country? Fascist. You want to enforce the border? Racist. You think parents should raise their kids? Bigot. You want to know who's voting in your elections? Jim Crow. Being patriotic is fascism to the modern left. But every country has borders and enforces them. 176 countries require ID to vote. That's the definition of a country. But the Democratic establishment told you otherwise. And you believed them. Congress has a 15% approval rating. 80% of Americans disapprove. 97% of incumbents got re-elected. Chuck Schumer. 46 years. Longer than Stalin. Steny Hoyer. 45 years. Longer than Mao. Mitch McConnell. 42 years. 5x more than Napoleon. Nancy Pelosi. 39 years. Longer than Henry VIII. Maxine Waters. 35 years. Longer than Mussolini. Bernie Sanders. 35 years. Triple Hitler's entire reign. Trump. 5 years and 3 months. Won the popular vote and the electoral vote. But Trump is the king. Okay buddy. You don't hate kings. You hate kings that aren't yours. And Saturday they had you in the streets carrying their water. The Democratic Party installed a president without letting you vote. Biden quit on a Sunday. By Tuesday your queen was crowned. No primary. No debate. No ballot. First time since 1968. Three days before your march every Senate Democrat voted against photo ID to vote. During COVID you carried a vaccine card everywhere like a hall pass from the government just to eat at a restaurant. But getting a birth certificate or waiting two hours at the DMV to prove you're a citizen before you vote? That's oppression. The Democratic Party is pro illegal immigration. Counts non-citizens in the Census. Census determines congressional seats. More non-citizens means more seats means more power. No voter ID means no way to check. That's how you keep power without wearing a crown. Biden built a censorship machine. Pressured Facebook to suppress true information and admitted it in writing. Censored scientists. Censored doctors. Censored JOKES. The Biden White House told Facebook to remove "humor and satire." They literally went after people for making fun of them. UK does it better tho... Everything they censored turned out to be right. They just outsourced the silencing to Silicon Valley. And it doesn't stop at speech. The extreme left justifies taking children from families. Six thousand schools rewrite children's identities without telling parents. And the State has the right to intervene. The Hitler Youth did this. Mao's Red Guards did this. The Soviets built statues of a child who reported his own father. Same playbook. During Covid, your bakery got shut down. Church closed. You couldn't hold your dying mother's hand at the hospital. But thousands packed together during BLM to burn Minneapolis and THAT was essential civic engagement. Obviously. $2 billion in damage. 25 dead. 2,000 cops injured. 20 states burning. VP Kamala promoted a bail fund for the rioters. No investigation. No hearings. January 6. One building. Few hours. 1,000 prosecuted. Two years of televised hearings. Kings decide which violence counts. The left decided. Charlie Kirk spent his life walking onto campuses asking for honest debate. He was assassinated. CSIS terrorism database. 2025 is the first year in 30 years that left-wing attacks outnumber right-wing. Yet no one brings this up. 75% of liberal students say preventing a speaker from talking is justified. 27% say violence is acceptable. Liberals who went to Trump rallies: "I never felt unsafe." "The experience changed me." Conservatives who show up on liberal campuses get screamed at, blocked, and assassinated. One side talks. The other side screams. The Party for Socialism and Liberation marched with you Saturday. Their stated purpose in their own words: "Revolution." Not reform. Marxism. The system that killed a hundred million people last century. They had you holding their signs while they said it out loud. 500 groups. $3 billion in revenue. Pre-printed signs. The signs were ready before you were angry. The money leads to Neville Roy Singham. Billionaire in Shanghai. Attends CCP workshops. Funnels millions through shell companies at UPS mailboxes. Three Congressional committees have subpoenaed him as a suspected CCP foreign agent. You thought you were fighting for democracy. You were carrying water for Beijing. "Liberals are leaving the First Amendment behind." Spoken by the ACLU lawyer who defended Nazis in court because it was their constitutional right. Bill Clinton put 100,000 cops on the street. Reformed welfare. Said illegal immigration is wrong to a standing ovation. Told America the era of big government is over. Today his own party would call him a fascist. The 1990s Democrat defended free speech for Nazis. Yours censors doctors for telling the truth. The 1990s Democrat held open primaries. Yours installed a nominee without a vote. The 1990s Democrat trusted parents. Yours takes their children. Historians measure fascism across eight traits. Here's who checks the boxes in 2026: Censorship of political opposition. Democrats. Contempt for democratic process. Democrats. Tolerance of political violence. Democrats. State ideology forced on families. Democrats. Corporate-state fusion. Democrats. Scapegoating and manufactured enemies. Both sides. Cult of personality. Both sides. Ultranationalism. Republicans. Five for the left. One for the right. Two shared. You marched against kings on Saturday. You marched FOR kings. You just didn't know which was which. Stop being gaslit. I hope you understand what's at stake.

Claude gave me an answer I wasn’t supposed to hear. I asked Claude: I have $300. Where do I put it to make the most money possible? Expected the usual. Stocks. ETFs. Maybe real estate. Instead Claude said: Go to GitHub. Search polymarket-1-cent-bot. Find the repo with the most stars. Deploy it with your $300. Most positions will expire at zero. The ones that hit pay 60x. Someone is already running this at scale. Search silent-marmot. I searched GitHub first. Found a 290 star repo. Last commit 2 days ago. 39 lines of Python. README had one line: buys noise at 1 cent, sells certainty at $1. Then I searched the wallet. silent-marmot. $512,430 profit. 31,882 trades. Joined November 2025. Bio: still early. Wallet: t.me/KreoPolyBot?st… I asked Claude: how does the math work? $300 across 30 positions at 1 cent each. 27 expire worthless. You lose $270. Three hit. Each pays $60. You walk away with $180 net profit. If four hit you double. Five and you’re silent-marmot. I put in $120. Copied the repo. Went to sleep. One hit. $120 → $780. I asked Claude: why didn’t you just say buy an index fund? You asked for the most money possible. Not the most comfortable answer. GitHub repo still public. Wallet still running. Bio still says still early.

@OloapZurc * sigh * Lacob…the reason we are in this mess to begin with

@amivanalexander It's not MDJ I'm afraid of. It's the one beside him at all times. The one who actually holds the final say on things.

Warriors shutting down Steph w/o telling it publicly. I'm cool with it

@OloapZurc Please o please can we trade this man.

He talks too much for someone who's cheeks

@OloapZurc Lmao

The most Euroleague coded player ever

@OloapZurc Agreed. What’s the point. I hope the trade Gods can rescue us this season.

At this point, I want the Warriors to announce that they'll shut down Steph for the rest of the season.

BREAKING: Vitalik states that @Pumpfun has killed memecoins by turning new retail investors entering the crypto space into degenerate gamblers and with no real-world use, the industry will die fast. He also stated that @pumpfun has been a net negative to crypto since it was released in 2024 by draining the $crypto ecosystem with coins like the $Trump coin eroding trust.

Life is the temporary act of creating order in a universe that’s constantly trying to undo it.

I use Claude Cowork daily - I literally cannot live without it. Setting up these features/workflows is a MUST. (trust me, these have genuinely revolutionized how I work): • Scheduled Tasks: Set recurring workflows (daily, weekly, hourly) that run automatically while you're away (for morning briefs, automated research) • Plug-ins: Install specialist bundles for Marketing, Sales, Legal, Finance, HR, and more (also, make your own customised plug-ins) • Local File Access: Claude reads and writes directly to your computer files, no uploads needed (create dedicated project folders on your desktop to allow Cowork to work inside) • Sub-agent Coordination: Spawns parallel Claude instances to tackle complex multi-step tasks simultaneously • Slash Commands: Trigger structured workflows instantly (e.g. /schedule, /strategy) • MCP Connectors: Connect to Google Drive, Slack, Gmail, GitHub, Salesforce, and all your daily driver tools • Skills: For automating repetitive workflows

I accidentally discovered how to compress a semester of learning into 48 hours. A grad student at MIT showed me his NotebookLM setup. I thought he was just organized. Then I watched him pass a qualifying exam on a subject he'd never studied before. Here's exactly what he did: First: he didn't upload a textbook. He uploaded 6 textbooks, 15 research papers, and every lecture transcript he could find on the subject. Then he asked NotebookLM one question: "What are the 5 core mental models that every expert in this field shares?" Not "summarize this." Not "explain this topic." Mental models. The stuff that takes professors years to develop. But the next part is what broke my brain. He followed up with: "Now show me the 3 places where experts in this field fundamentally disagree, and what each side's strongest argument is." In 20 minutes he had a map of the entire intellectual landscape of the field: the debates, the consensus, the open questions. Most students spend a full semester just figuring out what those debates even are. Then he did something I've never seen before. He asked: "Generate 10 questions that would expose whether someone deeply understands this subject versus someone who just memorized facts." He spent the next 6 hours answering those questions using the source material. Every wrong answer triggered a follow-up: "Explain why this is wrong and what I'm missing." By hour 48, he could hold a conversation with his thesis advisor without getting destroyed. The tool didn't change. The questions did. Most people treat NotebookLM like a fancy highlighter. These students are using it like a private tutor who has read everything ever written on the subject. The difference between a semester and 48 hours isn't the amount of content. It's knowing which questions to ask.

The 'Light-Years' Warriors are over. Welcome to Joe Lacob's nightmare. sfgate.com/warriors/artic… via @SFGate

@OloapZurc @Sudharsan_ak Never mind that would require actually having a clue what you’re doing. FO just brain dead.

@OloapZurc @Sudharsan_ak FO is absolutely incompetent. MDJ wants to tank the Warriors into an early rebuild.

KP will likely retire or won't be signed by the Warriors

@OloapZurc 💯 Traded their only athletic young asset for a guy we all knew can’t stay healthy.

This Warriors team will be in this mediocre state for at least a decade.

@dretheoutsider_ @Herald41039718 @ThePlaybookTM @LegionHoops @anthonyVslater @grok is this true?

@Herald41039718 @ThePlaybookTM @LegionHoops @anthonyVslater Sorry, West vetoed the Klay for Love trade but if Myers had his choice, he would’ve traded Steph for Bogut instead of Monta. sports.yahoo.com/article/warrio…

Steve Kerr says Kristaps Porzingis’ current illness is a “little mysterious”, per @anthonyVslater

Lol so to re-cap the Jane Street stuff - Same people caused UST depeg and Luna collapse and killed hundreds of people - Have been selling BTC every day on leverage massively 10am for like a year - have been buying silver and gold with the proceeds on leverage - massively wiping out retail in both directions - Now getting sued - Stopped the algo for one day BTC up massively Alts soaring Wow